How to handle your own preprints when prepping a camera-ready

Writing the camera-ready version of a paper that was already living on ArXiv can feel like a weird academic paradox. You've hit that stage where your work has already started circulating, people are citing your preprint, and others are already building on top of your methodology. Now that you have the official conference acceptance, you’re staring at the Related Work section wondering if you should cite yourself, how to acknowledge the momentum your preprint created, and how to ensure you don't accidentally make your own work look "old" or unoriginal.

The anxiety usually stems from two fears: looking like you're self-promoting by citing your own preprint, or accidentally diminishing your own novelty by pointing out that others are already doing similar things.

The citation dilemma

First off, you absolutely can—and should—cite your preprint if it contains foundational details or specific technical derivations that the camera-ready version expands upon. In the machine learning community, ArXiv is the standard. If a reader wants to see the full experimental setup that didn't make the final page limit, pointing them to the preprint is helpful, not arrogant.

However, there is a specific way to handle the "self-citation" aspect to avoid looking like you're padding your metrics. Instead of just dropping a citation, use it to provide context. For example, if the conference version is a condensed or refined version of the preprint, you can include a footnote on the first page:

\thanks{A preliminary version of this work appeared in [Preprint Citation/ArXiv ID].}

This is standard practice and tells the reviewers and readers exactly what the relationship is between the two documents.

Addressing subsequent works without losing novelty

This is the trickier part. If you see papers coming out that use your methodology, it's a double-edged sword. On one hand, it proves your work is impactful and relevant. On the other, if you describe them poorly, it might look like your work was just a stepping stone rather than a primary contribution.

The best approach is to frame these subsequent works as validation rather than competition. In your Related Work or Discussion section, you can position your work as the baseline that has already catalyzed new directions.

Instead of saying:
"Work X also uses our method to solve Y,"

Try something like:
"Our proposed methodology has recently been adopted by several studies to address [specific problem] (e.g., [Citation]), demonstrating its versatility in [specific domain]."

This phrasing does three things:
1. It asserts ownership of the methodology.
2. It shows your work is a recognized standard in the field.
3. It prevents the "scooping" feeling by framing the other authors as users of your innovation rather than competitors of your idea.

Protecting your technical contribution

When you're deep in the prompt engineering or architecture design phase of your paper, you want to ensure your specific "delta"—the actual new thing you brought to the table—is crystal clear. If subsequent papers are extending your work, use that as an opportunity to highlight your original contribution.

If Paper B extends your method, you can write:
"While [Paper B] extends our framework to include [new feature], our original work provides the fundamental [core mechanism] necessary for [primary task]."

By doing this, you aren't just listing papers; you are building a narrative of your own research's influence. You're turning a potential "novelty threat" into a "proven impact" metric. It shows the program committee that your work isn't just a theoretical exercise—it's already being integrated into the real-world AI workflow of other researchers.